Different molecular bases underlie the mitochondrial respiratory activity in the homoeothermic spadices of Symplocarpus renifolius and the transiently thermogenic appendices of Arum maculatum

Organ-specific thermogenesis associated with floral development has been well documented in plants from a diverse range of families, including the monocotyledonous Araceae [1], dicotyledonous Nelumbonaceae [2] and gymnospermous Cycadaceae [3]. Most thermogenic activities in these plants can be categorized into two types, in accordance with the transition patterns of their tissue temperature. Transient heat production has been observed in the majority of thermogenic plants analysed, where the temperature of a thermogenic organ rises discontinuously over one or two events. This type of thermogenesis has been well studied in Arum species since Jean-Baptiste Lamarck first documented the thermogenic activity of plants in the European arum lily [4]. For instance, it is now widely known that appendices of Arum maculatum can be hotter than 30°C at the peak of their transient thermogenesis, when the ambient temperature is between 7–22°C [5,6,7]. On the other hand, homoeothermic heat production has also been reported in a limited number of species, such as sacred lotus (Nelumbo nucifera) and skunk cabbage (Symplocarpus species) [2,8,9]. In these plants, the thermogenic organ remains at a constant temperature for several days with remarkable precision. For example, the spadices of S. renifolius are reported to maintain their tissue temperature at a relatively stable 21±2.4°C, although the ambient temperature varies over a wide range [8]. Surprisingly, this plant can produce heat that is sufficient to melt the surrounding snow in its natural environment [10].

The molecular basis of thermogenesis in plants has long been associated with the expression of cyanide-resistant respiration [4]. Such respiration is attributed to the activity of AOX (alternative oxidase), which is located on the matrix surface of the inner mitochondrial membrane and catalyses the oxidation of ubiquinol and the reduction of molecular oxygen to water [11]. AOX is an energy-dissipative enzyme, since it does not contribute to the formation of a proton electrochemical gradient across the inner membrane and, as a consequence, to ATP synthesis. The drastic change in free energy which is accompanied by the reaction catalysed by AOX is believed to be liberated as heat in thermogenic plants. In fact, a large cyanide-resistant respiration capacity has been observed to be associated with thermogenesis in plants, regardless of whether it is transient or homoeothermic [12,13]. Although a large quantity of AOX protein has been shown to accumulate in mitochondria following the onset of thermogenesis in some species [13,14], this is not the prerequisite for a considerable cyanide-resistant respiration capacity as previously demonstrated in A. maculatum [12].

BN-PAGE (Blue native PAGE) has become a major technique in the analysis of multiprotein complexes, particularly those localized to membranes [15]. This method is especially effective when combined with 2D (two-dimensional) SDS/PAGE and MS, as it is then possible to identify the constitutive subunits of a complex. All of the typical plant mitochondrial respiratory complexes have been characterized using this technique [16,17], the major findings of which have revealed that there are distinct differences between plant and mammalian mitochondria, such as an eight-subunit complex II and also plant-specific subunits that are associated with complex IV [18]. A number of the supramolecular structures of the respiratory complexes, i.e. supercomplexes, have also been successfully characterized using BN-PAGE in plants. Supercomplex I+III₂ is highly abundant in a number of species [16], whereas the expressions of other supercomplexes, such as I₂+III₄ and I+III₂+IVa₄, which are generally at low abundance, vary even between organs of the same species [17].

Previously, the structures of the respiratory complexes in mitochondria from thermogenic appendices of A. maculatum were also analysed using BN-PAGE, but few striking characteristics that are unique to thermogenic species were detected [19]. In view of the finding that that study only focused on A. maculatum, but did not consider homoeothermic plants, which appear to precisely regulate their metabolic flux, it is conceivable that such tissues may possess different respiratory characteristics in comparison with A. maculatum. Furthermore, although AOX has been analysed in previous studies, scant attention has been paid to other non-oxidative phosphorylation components, such as the type II NAD(P)H dehydrogenases. In light of the above, further investigations into the characteristics of the respiratory chain of thermogenic plants seems warranted.

In the present study, we have analysed the expression profile and native structures of the respiratory components of mitochondria from the stigma-stage spadix of S. renifolius, a homoeothermic organ, in addition to the δ-stage appendix of A. maculatum, a transiently thermogenic organ. Our primary objective was to examine whether mitochondria of these plants have different molecular mechanisms underlying their respective thermogenic activities. To this end, not only the respiratory components, but also an extensive range of mitochondrial proteins, were subjected to qualitative comparative analyses between these two species. We reveal two notable differences in the respiratory chains of these plants; namely an irregularly large sized complex II is present in S. renifolius, but not in A. maculatum, and a type II NAD(P)H dehydrogenase appears to be expressed at a much higher level in A. maculatum than in S. renifolius. Furthermore, our results suggest that the respiratory pathways which normally function during thermogenesis are different in the two thermogenic plants.

For the preparation of mitochondria, spadices of S. renifolius were collected from their natural environments in Hakuba, Nagano prefecture, Japan, in April 2008, as well as in Omori, Akita prefecture, Japan, in April 2010. Appendices of A. maculatum were collected from the campus of the University of Sussex, Brighton, U.K., in May 2010. The thermogenic stages of the plants were classified according to the morphological changes in their inflorescences as described previously [5,9,20]. For the preparation of RNA, tissue samples of S. renifolius were collected in Daisen, Akita prefecture, Japan, in March 2007, and in Fujine, Iwate prefecture, Japan, in March 2012, and those of A. maculatum were collected on the campus of the University of Sussex in May 2009. These samples were flash-frozen in liquid nitrogen to prevent RNA degradation.

Mitochondria from S. renifolius spadices and A. maculatum appendices were isolated as described previously [21,22]; however, 27%–45%–60% Percoll gradients were used for both S. renifolius and A. maculatum. Mitochondria from Schizosaccharomyces pombe Sp.011 were isolated according to an established protocol [23]. The isolated mitochondria were stored at −80°C and were aliquoted on the first thawing.

Protein concentrations of isolated mitochondria were determined by the BCA (bicinchoninic acid) method using a BCA Protein Assay Kit (Pierce). Various concentrations of BSA in the range of 0–2000 μg/ml were used as the reference standards.

Mitochondrial proteins were solubilized in digitonin using a NativePAGE Sample Prep Kit (Invitrogen) as instructed by the manufacturer. The 1D BN-PAGE was performed using NativePAGE Novex 4–16% Bis-Tris Gels (Invitrogen) and an XCell SureLock Mini-Cell (Invitrogen). Approximately 50 μg of the solubilized proteins was loaded per lane and electrophoresed for 150 min at 150 V with NativeMark Unstained Protein Standards (Invitrogen). After 50 min of electrophoresis, ‘dark cathode buffer’ (Invitrogen) containing 0.02% Coomassie Brilliant Blue G-250 was substituted with ‘light cathode buffer’ (Invitrogen) containing 0.002% Coomassie Brilliant Blue G-250. For 2D BN/SDS/PAGE, a gel strip underwent reductive alkylation as instructed by the manufacturer and was then loaded on to NuPAGE Novex 4–12% Bis-Tris ZOOM Gels (Invitrogen). The electrophoresis was performed in NuPAGE Mops SDS Running Buffer (Invitrogen) for 55 min at 200 V. PageRuler Prestained Protein Ladder (Fermentas) or Precision Plus Protein Dual Color Standards (Bio-Rad Laboratories) was used as the protein standards for the 2D BN/SDS/PAGE. Visualization of the proteins was achieved by incubating the gels in Quick-CBB PLUS (Wako Pure Chemical Industries).

Protein spots visualized on a 2D BN/SDS/PAGE gel were excised with sterile knives and were subjected to in-gel digestion with Sequencing Grade Modified Trypsin (Promega). The resulting peptides were purified using Zip Tip pipette tips (Millipore) and were eluted in 15 μl of 0.1% TFA (trifluoroacetic acid). The peptide solutions were then separated using an Advance UHPLC (ultra-high performance liquid chromatography) system (Michrom Bioresources) connected to an L-column Micro 0.3 mm×5 mm (CERI) as the trap column and a Magic C18 AQ nano column 0.1 mm×150 mm (Michrom Bioresources) as the separation column. The retained peptides were eluted using a linear gradient of acetonitrile from 5% to 45% (v/v) at a flow rate of 500 nl/min and were subsequently ionized with an Advance spray source (Michrom Bioresources) at 1.8 kV. Mass spectrometric analyses of the peptides were carried out using an LTQ Orbitrap XL (Thermo Fisher Scientific) operated by Xcalibur software version 2.0.7 (Thermo Fisher Scientific). The raw MS/MS (tandem MS) spectra files were first converted into mgf (Mascot generic format) files using Proteome Discoverer (Thermo Fisher Scientific) and were subsequently processed by MASCOT search version 2.3.02 (Matrix Science) with the following parameters: database, NCBInr + contaminants (compiled by the Max-Planck Institute of Biochemistry, Martinsried, Germany); arrow up to 1 missed cleavage; taxonomy, viridiplantae; variable modifications, oxidation of methionine; peptide mass tolerance, 10 p.p.m.; MS/MS tolerance, 0.2 kDa; peptide charge, 2⁺ and 3⁺; significance threshold, P<0.05; and cut-off expectation value, 0.05.

Proteins separated by either 1D BN-PAGE or 2D BN/SDS/PAGE were transferred to an Immobilon P membrane (Millipore) at 45 V and 1 mA per cm² for 90 min. In the case of Western blotting following 1D BN-PAGE, the membrane was sequentially incubated in 8% acetate for 15 min and then in deionized water for 5 min to be dried; this drying procedure was indispensable for clear immunological detection because methanol incubation during the rehydration process eliminates undesirable Coomassie Brilliant Blue G-250 from the membrane. The membrane was then treated as described previously [21] with the addition of an overnight incubation with AOA monoclonal antibody (Agrisera) at 4°C.

Total RNAs were extracted from thermogenic organs, i.e. a stigma-stage spadix of S. renifolius and a γ–δ-stage appendix of A. maculatum, using either an RNeasy Plant mini kit (Qiagen) or a FastPure RNA kit (Takara Bio). First strand cDNAs were generated using a PrimeScript II 1st strand cDNA Synthesis kit (Takara Bio) and an oligo dT primer. Partial fragments of the targeted transcripts were first amplified by PCR using the following primers (for all primer sequences, see Supplementary Table S1 at http://www.BiochemJ.org/bj/445/bj4450237add.htm): NDAF1-NDAR1, NDAF2-NDAR2 and NDAF3-NDAR3 for fragments from NDA; NDBF1-NDBR1, NDBF2-NDBR2 and NDBF3-NDBR3 followed by NDBF4-NDBR4 for fragments from NDB; and EF1aF1-EF1aR1 for fragments from EF1α. Gene-specific primers were then prepared to perform 5′- and 3′-RACE (rapid amplification of cDNA ends) with the SMARTer RACE cDNA Amplification kit (Clontech). The primers were used as follows: SrNDAR1 (where Sr is S. renifolius) for 5′-RACE of SrNDA; SrNDAF1 for 3′-RACE of SrNDA; SrNDBR1 and SrNDBR2 for 5′-RACE of SrNDB; SrNDBF1 and SrNDBF2 for 3′-RACE of SrNDB; AmSrEF1aR1 (where Am is A. maculatum) and AmSrEF1aR2 for 5′-RACE of SrEF1α; AmSrEF1aF1 for 3′-RACE of SrEF1α; AmNDAR1 for 5′-RACE of AmNDA; AmNDAF1 for 3′-RACE of AmNDA; AmNDBR1 for 5′-RACE of AmNDB; AmNDBF1 for 3′-RACE of AmNDB; AmSrEF1aR1 and AmSrEF1aR2 for 5′-RACE of AmEF1α; and AmSrEF1aF1 for 3′-RACE of AmEF1α. To isolate the full-length cDNAs, final PCR amplifications were performed with KOD-Plus (Toyobo) using the following gene-specific primers: SrNDAF2-SrNDAR2 for SrNDA1 (DDBJ accession number: AB677831) and SrNDA2 (AB677832); SrNDBF3-SrNDBR3 for SrNDB1 (AB677833) and SrNDB2 (AB677834); SrEF1aF1-SrEF1aR1 for SrEF1α1 (AB677835); AmNDAF2-AmNDAR2 followed by AmNDAF3-AmNDAR3 for AmNDA1 (AB677836); AmNDBF2-AmNDBR2 for AmNDB1 (AB677837), and AmEF1aF1-AmEF1aR1 for AmEF1α1 (AB677838). The obtained fragments were then cloned into pCR2.1 (Invitrogen) and sequenced in both directions. The large (AG)_n region and its proximity in the 5′-UTR (untranslated region) of AmNDB1 in some cases enabled it to be sequenced from one direction only. Each full-length sequence was determined by isolating two clones carrying identical inserts. However, slight differences in the number of simple sequence repeats such as (AG)_n and (CT)_n within the UTR were allowed for. DNA sequences were analysed using GENETYX (Genetyx).

The phylogeny of the type II NAD(P)H dehydrogenases from S. renifolius and A. maculatum was analysed using MEGA4 [24] at the protein level. An alignment was generated using a ClustalW algorithm with the following sequences (GenBank® accession numbers in parentheses): Arabidopsis thaliana NDA1 (NP_563783), AtNDA2 (NP_180560); Oryza sativa NDA (NP_001044694), OsNDA (NP_001060003); Populus trichocarpa NDA (XP_002305005), PtNDA (XP_002313376), PtNDA (XP_002317236); Ricinus communis NDA (XP_002521423), RcNDA (XP_002525658); Sorghum bicolor NDA (XP_002456580), SbNDA (XP_002460866); Solanum tuberosum NDA (CAB52796); Vitis vinifera NDA (XP_002262771), VvNDA (XP_002277505); Zea mays NDA (NP_001168889); SrNDA1; SrNDA2; AmNDA1; AtNDB1 (NP_567801); AtNDB2 (NP_567283); AtNDB3 (NP_193880.5); AtNDB4 (NP_179673); OsNDB (NP_001055220); OsNDB (NP_001058394); PtNDB (XP_002305652); PtNDB (XP_002316663); PtNDB (XP_002319383); PtNDB (XP_002319384); PtNDB (XP_002319927); RcNDB (XP_002517990); RcNDB (XP_002517991); RcNDB (XP_002524685); SbNDB (XP_002438881); SbNDB (XP_002445027); SbNDB (XP_002456448); StNDB (CAB52797); VvNDB (XP_002274469); VvNDB (XP_002274523); VvNDB (XP_002274690); ZmNDB (NP_001142169); SrNDB1; SrNDB2; AmNDB1; AtNDC1 (NP_568205); and OsNDC1 (BAD35311). On the basis of the deduced alignments, the phylogeny of the proteins was inferred using the minimum evolution method, and a rooted phylogenetic tree was obtained by setting the NDC cluster as an out-group. The reliability of the tree was examined using the bootstrap test with 1000 replicates, and branches with partitions reproduced in less than 50% replicates were collapsed.

cDNAs were first generated using a ReverTra Ace qPCR RT (reverse transcription) Kit (Toyobo), and quantitative real-time PCR was then performed with SYBR Green Real-time PCR Master Mix-Plus (Toyobo) using a Thermal Cycler Dice TP800 (Takara Bio). Gene-specific primers were designed against the consensus sequences for both a targeted transcript in A. maculatum and its homologue(s) in S. renifolius, and used as follows: rtNDAF1-rtNDAR1 for the NDA transcripts; rtNDBF1-rtNDBR1 for the NDB transcripts; and rtEF1aF1-rtEF1aR1 for the EF1α transcripts. PCR was performed using the following conditions: an initial denaturing step at 95°C for 10 s; 60 amplification cycles of 95°C for 15 s, 57°C for 15 s and 72°C for 30 s; and a dissociation curve analysis at 95°C for 15 s, 60°C for 30 s and 95°C for 15 s. Significant differences between the means were examined by one-way factorial ANOVA and the Tukey's honestly significant difference post-hoc test using SPSS software (IBM).

To characterize the expression profiles of the mitochondrial respiratory complexes in the thermogenic plants S. renifolius and A. maculatum, mitochondria from stigma-stage spadices of S. renifolius and those from δ-stage appendices of A. maculatum were isolated. The mitochondrial proteins were then resolved using 1D BN-PAGE after solubilization with various concentrations of digitonin (Figure 1). As shown in the Figure, 1 g of digitonin per g of protein was not sufficient to solubilize either mitochondrial preparation. However, the use of 2.5 or 5 g of digitonin per g of protein worked well in both cases and did not result in different resolutions of the complexes by Coomassie Brilliant Blue staining, with the exceptions of complex V and complex III₂. These two complexes were visualized at slightly lower positions when using 5 g of digitonin. Since Eubel et al. [16] had demonstrated previously that 5 g of digitonin per g of protein is optimal for visualizing complexes in Arabidopsis mitochondria, a number of subsequent BN-PAGE studies in plants have employed this concentration. We adopted this dosage in the experiments described below to facilitate a more direct comparison between the results of the present study and previously reported data.

In addition to complex V and complex III₂, bands corresponding to supercomplex I+III₂, complex I, complex II and complex II Fp (flavoprotein) subunits were also recognizable on the gel (Figure 1). The identities of the respective bands were first predicted by a comparative approach on the basis of previously published results [16,17] and were later confirmed by gel-based proteomic analysis, as described below. On the whole, the expression and mobility of most of the respiratory complexes on the 1D BN-PAGE gels were very similar between the mitochondria from the two thermogenic plants. The only exception to this was complex II, which in S. renifolius mitochondria had drastically less mobility, suggesting a difference in its molecular size. Although it has been widely reported that the mobility of a protein does not correlate well with its molecular mass on a native gel [17], it should be noted that complex II of S. renifolius and that of A. maculatum had mobilities of 340 kDa and 120 kDa respectively.

To further investigate the structures of the respiratory complexes in more detail, their constitutive subunits were separated by 2D BN/SDS/PAGE (Figure 2) and the protein spots visualized by Coomassie Brilliant Blue staining were further analysed by MS to determine their molecular identities (Table 1). It appears that four known subunits of complex I, namely the 75 kDa, 49 kDa, 39 kDa and 27 kDa proteins, were identified in a lane that corresponds to the position of complex I in the first gel dimension (Figures 2A and 2B). Likewise, four subunits were identified in a lane for complex V; five subunits in a lane for complex III₂; and seven subunits derived from complex I and complex III in a lane for supercomplex I+III₂. Heat shock protein 60-like proteins were also detected in lanes for complex V in the mitochondria from both thermogenic plants, as reported previously in yeast mitochondria [25]. Complex II could be characterized by Fp subunits and, when available, iron-sulfur subunit (Ip subunit) on the gels. No notable differences were found in the composition of the respiratory complex subunits between S. renifolius and A. maculatum mitochondria.

Coomassie Brilliant Blue staining visualized approximately 70 2D gel protein spots, regardless of the origin of the mitochondria (Figure 2), and most of these visible spots were subjected to protein identification by MS. We thereby succeeded in identifying approximately 50 proteins for each mitochondrial sample (Figure 2 and Table 1). A total of 44 of these proteins were comparable between the S. renifolius and A. maculatum mitochondria in terms of their positions on the gels. All of these consensus proteins turned out to be mitochondrial with only one exception, cytosolic aconitase. This is strong evidence that the mitochondria we had prepared were highly purified. A total of 5 proteins were identified as specific for S. renifolius mitochondria (A–E in Figure 2 and Table 1). With respect to mitochondrial respiratory components, these included succinate dehydrogenase subunits, which reflect the small mobility of S. renifolius complex II in the first gel dimension, and an uncoupling protein [26].

A total of seven proteins were identified as specific for A. maculatum mitochondria (a–g in Figure 2 and Table 1). For reasons similar to the case of S. renifolius complex II, succinate dehydrogenase Fp subunit was included in this list. Type II NAD(P)H dehydrogenase, which is found on the external face of the mitochondrial inner membrane (NDB) and AOX were also identified as A. maculatum mitochondria-specific proteins. The spot corresponding to NDB was clearly recognizable by 2D gel analysis of A. maculatum mitochondria (Figure 2B), but appeared to be absent on the S. renifolius mitochondria gel (Figure 2A). In the case of AOX, this was also identified as a consensus protein from spot number 13, and the A. maculatum-specific identification of this protein was due to its additional distribution on the 2D gel.

Since the expression status of NDB was found to be different between S. renifolius mitochondria and A. maculatum mitochondria (Figure 2), we isolated cDNAs encoding type II NAD(P)H dehydrogenases from the thermogenic organs of these two species and analysed the expression levels of these transcripts within the cDNA pools. Using RT–PCR, a cDNA encoding a type II NAD(P)H dehydrogenase found on the internal face of mitochondrial inner membrane (NDA) and a further cDNA encoding NDB were isolated from each plant. The translation initiation codons of the respective transcripts were predicted on the basis of the consensus sequence reported previously [27], and the deduced amino acid sequence of each transcript was generated (Supplementary Figures S1 and S2 at http://www.BiochemJ.org/bj/445/bj4450237add.htm). AmNDA1 and two SrNDAs were theoretically determined to be 61 kDa proteins in their precursor forms. However, they were predicted to form 54 kDa mature proteins after the putative mitochondrial targeting sequences predicted by Mitoprot were cleaved (Supplementary Figure S1). These predicted molecular masses are in good agreement with the 2D mobilities found by 2D BN/SDS/PAGE in which the NDA protein bands always appeared at approximately 54 kDa (Figure 2). AmNDB1 and two SrNDBs were predicted to be 65 kDa proteins in their precursor forms. Mitoprot further predicted that, although AmNDB1 would be processed into a 61 kDa protein, SrNDB1 and SrNDB2 would be processed into 63 kDa proteins (Supplementary Figure S2). Molecular mass estimates for the NDB protein spot from A. maculatum mitochondria was consistently determined to be 62 kDa (Figure 2B).

When aligned with previously characterized type II NAD(P)H dehydrogenases from potato, several functional motifs were identified in the dehydrogenases from S. renifolius and A. maculatum. Two dinucleotide-fold fingerprints, which have been reported to be well conserved in both NDAs and NDBs, were found in the amino acid sequences of the newly identified sequences (Figure 3A) [28,29]. In addition, the Ca²⁺-binding EF-hand motif first identified in potato NDB [28] was found to be conserved among the NDB sequences in the present study (Figure 3B). Another EF-hand motif reported previously in Neurospora crassa p64, a homologue of plant NDB, was not found in any of the NDB sequences [30]. Phylogenetic analysis of 15 plant NDA sequences, 22 plant NDB sequences and two plant NDC sequences clearly demonstrated the identities of the dehydrogenases from S. renifolius and A. maculatum as they were grouped into their respective protein families (Figure 3C). Intriguingly, the distribution of monocot- and dicot-derived dehydrogenases within this phylogenetic tree suggested the possibility that at least two subfamilies exist in each protein family.

The expression levels of transcripts encoding NDA and NDB were investigated in a spadix of S. renifolius and an appendix of A. maculatum, as well as in leaves and spathes of both species (NDA in Figure 3D and NDB in Figure 3E). Since this analysis could not discriminate between two homologous SrNDAs nor the two SrNDBs, the quantities of NDA and NDB in the samples from S. renifolius represent the gross expression levels of the two homologues in the respective gene families. Figure 3(D) reveals that NDA was expressed to the same degree in the two thermogenic organs. In non-thermogenic organs, however, NDA expression levels tended to be lower. On the other hand, NDB showed a significantly higher expression in the appendix of A. maculatum than in the spadix of S. renifolius. Furthermore, expression levels of NDB in the spadix and the spathe were comparable in S. renifolius. In the other non-thermogenic organs, the expression levels of NDB tended to be much lower.

Since AOX has been proposed to function in both homoeothermic thermogenesis in S. renifolius [21] and in transient thermogenesis in A. maculatum [7,12], its native structure in each plant was investigated. On a 1D image, AOX was detectable as a broad smear with a densitometric peak at approximately 200 kDa in both species (Figure 4A), the appearance of which was not largely affected by digitonin concentration or freeze–thaw treatments at least in S. renifolius (Supplementary Figure S3 at http://www.BiochemJ.org/bj/445/bj4450237add.htm). Mitochondria from non-transformed S. pombe and those from its transformant expressing AmAOX1e, the most abundantly expressed AOX in the appendices of A. maculatum [7], were also analysed as negative and positive controls, respectively (Supplementary Figure S4 at http://www.BiochemJ.org/bj/445/bj4450237add.htm). No signal was detected in the former, but a signal at approximately 200 kDa appeared very clearly in the latter.

AOX was detected as a horizontally broad smear at approximately 35 kDa in the 2D analysis, and in addition a minor smear at approximately 70 kDa was also recognizable (S. renifolius in Figure 4B and A. maculatum in Figure 4C). We suggest that this minor smear is derived from an oxidized form of AOX that had not been reduced sufficiently by reductive alkylation of a 1D gel strip. Obscure distribution of AOX in 2D BN/SDS/PAGE has previously been reported in non-thermogenic plants, namely Arabidopsis and Phaseolus vulgaris [16], as well as in thermogenic A. maculatum [19]. In the former report, the smear-like signals appeared at between 30 kDa and 300 kDa in the first dimension, whereas the latter study described signals of below 50 kDa. In contrast with the results of the latter report [19], our present analysis detected a smear-like signal of A. maculatum AOX in a much higher molecular mass region in the first dimension, which was quite similar to the observation made in the non-thermogenic plants [16]. This result was also supported by mass-spectrometric identification of the AOX protein on 2D BN/SDS/PAGE gels (Figure 2 and Table 1).

Some plants are known to undergo floral development accompanied by organ-specific thermogenesis. In Araceae, the stigma-stage spadices of S. renifolius were previously reported to show homoeothermic activity [8], whereas the δ-stage appendices of A. maculatum are known to raise their temperatures transiently [5]. Although increased respiratory activities via the energy-dissipative AOX have been shown to correlate with thermogenesis in both species [12,21], the molecular bases for their different thermogenic properties have not been well characterized. Hence, in the present study, we comparatively analysed the expressions and native structures of respiratory components within mitochondria isolated from the thermogenic organs of S. renifolius and A. maculatum.

In addition to four of the five typical respiratory complexes, namely complex I, II, III₂ and V, supercomplex I+III₂ was identified in the mitochondria from each plant (Figure 1 and Figure 2). This supercomplex has been reported to be expressed abundantly in various non-thermogenic plants, where 50%–90% of complex I is integrated into this supercomplex [16]. In the case of thermogenic plants, a previous study has reported that all detectable complex I forms a supercomplex in A. maculatum [19], whereas only 60% was found to do so from our present results. This inconsistency may derive from different storage conditions or different numbers of freeze–thaw cycles that the mitochondria had to experience, as the amount of some supercomplexes has been reported to be significantly affected by these factors [16]. In any case, the high abundance of supercomplex I+III₂, which probably contributes to the substrate channeling of the COX (cytochrome c oxidase)-mediated respiratory pathway, suggests that AOX mainly utilizes reducing equivalents from ubiquinone reductases other than complex I in S. renifolius and A. maculatum.

Besides complexes I and II, two NAD(P)H dehydrogenases are known to constitute the mitochondrial respiratory chain in plants which reduce ubiquinone to ubiquinol without any net proton translocation across the inner membrane [31]. In the present study, NDB, the NAD(P)H dehydrogenase expressed on the external surface of the inner membrane, was found to be abundant in the appendices of A. maculatum, but not in the spadices of S. renifolius (Figure 2). This was also confirmed at the transcript level (Figure 3E). Such results suggest that NDB supplies a great deal of reducing power from a cytosolic NAD(P)H pool to the mitochondrial respiratory chain or, more speculatively, to AOX in the appendices of A. maculatum. In this regard, high rates of glycolysis have been reported in the thermogenic spadices of A. maculatum in association with increased maximum catalytic activities of the glycolytic enzymes [32]. Moreover, it was previously reported that the oxygen consumption rate of AOX is 2.3-fold higher than that of COX in mitochondria isolated from spadices of A. maculatum when NADH was externally added as a respiratory substrate at 21°C [33]. Hence, the availability of exogenous NADH, which is probably linked to glycolytic activity in the cytosol, is an important factor underlying the high respiratory activity of AOX in this plant.

The other NAD(P)H dehydrogenase, the so-called NDA enzyme, was identified in the mitochondria from both plant species (Figure 2), and expression levels of the corresponding transcript(s) appeared comparable (Figure 3D). This suggests both S. renifolius and A. maculatum utilize this enzyme to provide reducing equivalents to the ubiquinone pool; however, S. renifolius, with less abundance of NDB, may depend more on NDA than A. maculatum. In S. renifolius, the difference between the floral and ambient temperature has been found to show a positive correlation with CO₂ production in the regulation of floral temperature [34]. As the CO₂ is presumably produced during TCA (tricarboxylic acid) cycle activity, the re-oxidation of NADH within the mitochondrial compartment is vital for maintaining cycle turnover, particularly at low ambient temperatures. Hence it is interesting to speculate that, in homoeothermic plants, perhaps NDA plays a more important role in the maintenance of TCA cycle activity.

Complex II is another crucial factor in the respiratory chain and in the TCA cycle. Interestingly, the molecular mass of this complex seems to be drastically different between the mitochondria from S. renifolius and those from A. maculatum (Figure 1). In A. maculatum, complex II had an apparent molecular mass of 120 kDa. This is similar to porcine complex II (124 kDa) [35] and Escherichia coli succinate:ubiquinone oxidoreductase (120 kDa), a prokaryotic homologue of mitochondrial complex II [36]. Both of them consist of four common subunits, but are slightly smaller than complex II of other plants [16]. Complex II has been reported to be 150 kDa in Arabidopsis, potato and P. vulgaris because they contain four additional plant-specific subunits, namely SDH5, SDH6, SDH7 and SDH8 [18]. Although homologues of these subunits seem to be conserved in monocots as well (results not shown), complex II of A. maculatum has been reported as a four-subunit complex instead of the eight-subunit complex described in a previous study [19]. Hence the cumulative evidence could be interpreted in two ways: (i) complex II of A. maculatum is indeed 120 kDa in vivo and the four plant-specific subunits are not necessary for its functionality; or (ii) complex II is actually 150 kDa in vivo, but appears as a 120 kDa complex on the gel owing to either partial dissection of the original form or a fast migration in native PAGE. In either case, complex II of A. maculatum probably exists in a monomeric form, as found in most organisms investigated to date.

Complex II of S. renifolius has an apparent molecular mass of 340 kDa. This is a very unusual size for this protein, in both plants and other organisms, because such a large molecular mass usually infers an oligomeric or a supramolecular structure. Thus far, however, complex II has never been shown to be an oligomer in vivo and has rather been reported as a monomer in yeast, mammals and plants [16,37]. The only report thus far that has shown evidence for an oligomeric structure of complex II was a report on the E. coli homologue, which suggested the complex may form a trimer under physiological conditions on the basis of its crystallographic packaging [36]. The supramolecular structure of complex II has not yet been elucidated either. However, several large versions of complex II have been detected weakly, but distinctly, by in gel activity staining of A. maculatum mitochondria [19]. Hence, a possibility that complex II exists as a trimer or a supercomplex in the spadices of S. renifolius cannot be ruled out at present and will require clarification in future studies.

No multiprotein complex comprising AOX has been identified to date, although the native structure of this energy-dissipative oxidase has already been analysed by BN-PAGE in Arabidopsis, P. vulgaris and A. maculatum [16,19]. However, the authors of these studies have inferred that AOX does not form a supercomplex and that smear-like signals of AOX in the first dimension are derived from aggregates of this enzyme. In A. maculatum, the smear was previously reported to appear below 50 kDa in the first dimension, but in the present study was detected in the 150–350 kDa ranges (Figure 4). AmAOX1e expressed in yeast also appears at approximately 200 kDa in the native condition (Supplementary Figure S4), suggesting that even in this heterologous system the same AOX-comprising complex is also apparent. In addition, the pattern of the smear was not significantly affected by the concentration of digitonin or freeze–thaw cycles in the case of S. renifolius AOX (Supplementary Figure S3). Therefore we speculate that the smear at approximately 200 kDa is not derived from an aggregation of the protein, but rather reflects a specific physiological situation.

In general, the expression profiles of the respiratory complexes, supercomplexes and other mitochondrial proteins that we investigated in our present study were very similar between S. renifolius and A. maculatum. However, several important members of the respiratory chain, such as NDB and complex II, were shown to have a different expression status between these plant species, and they may therefore play a crucial role in their respective thermogenic pathways, particularly in the regulation of electron flow to the respiratory chain. If this is indeed the case, such a key enzyme would be expected to regulate the metabolic flux with greater precision in the case of S. renifolius and to boost the flux up to its maximum capacity in the case of A. maculatum. In fact, the reduction level of the ubiquinone pool in the stigma-stage spadices of S. renifolius remains in the range of 40–75% [10], whereas that in the δ-stage appendices of A. maculatum reaches approximately 90% [5]. Hence it is important to study not only the native structures of these respiratory enzymes, but also the functional differences between species in order to properly elucidate the metabolic regulator in S. renifolius or the metabolic booster in A. maculatum.

1D

one-dimensional

2D

two-dimensional

AOX

alternative oxidase

BCA

bicinchoninic acid

BN-PAGE

Blue native PAGE

COX

cytochrome c oxidase

Fp

flavoprotein

MS/MS

tandem MS

RACE

rapid amplification of cDNA ends

RT

reverse transcription

TCA

tricarboxylic acid

UTR

untranslated region

Yusuke Kakizaki, Anthony Moore and Kikukatsu Ito designed the research. Yusuke Kakizaki conducted the experiments and analysed the data. Yusuke Kakizaki, Anthony Moore and Kikukatsu Ito wrote the paper.

We thank the members of our laboratories for technical assistance in the isolation of mitochondria as well as for helpful discussions. We also thank Daisuke Takahashi, Dr Matsuo Uemura and Dr Tetsuro Yamashita for their support with MS analysis.

This work was supported by a Grant-in-Aid for Scientific Research (B) from the JSPS (Japan Society for the Promotion of Science) [grant numbers 22405001 and 22380184]. Work in A.L.M.'s laboratory was supported by a grant from BBSRC (Biotechnology and Biological Sciences Research Council). Y.K. is a Research Fellow of the JSPS.